NAIRA: automatic processing of camera trap images

• 

  Manual Operation

• 

  Automatic Operation

• 

  Assisted Operation

Processing of Images obtained with Trap Cameras

Camera Traps are a non-invasive tool that allows to obtain photographs of wildlife animals that cross over their path. Thus, they are considered appropriate for the elaboration of inventories, estimates of abundance and density of vertebrate species of terrestrial habits. The use of this technology provides information on the presence and distribution of fauna of cryptic habit. However, it is necessary to spend long periods of time to process, identification of species and finally the construction of appropriate databases to perform the corresponding statistical analyzes.

The automatic processing of this kind of sequences are complex due to the constantly changes in the scenes, variations in light levels,different climatic conditions between triggers and other variations resulting from the natural dynamics of the ecosystem. These dissimilarities make difficult the automatic processing increasing the computational cost and the run time.

The software NAIRA was designed to help in the processing of camera trap images. Using machine learning techniques this tool is able to automatically identify the images with presence of fauna, to identify birds and mammals, and within this last group the taxonomic genus of the species present in each photograph. In addition, NAIRA automatically extracts metadata from photographs (moon phase, temperature, GPS coordinates, capture date and time) and allows the user to manually add information that he or she deems important for the analysis of trap camera images. All information generated in NAIRA is registered in the Darwin Core format for species registration.

Currently, NAIRA is trained to identify 20 important genus with distribution in the neotropic. The next step is to extend identification coverage to the worldwide wildlife.

NAIRA is the result of a joint effort between the SISTEMIC group of the University of Antioquia and the Alexander von Humboldt Biological Resources Research Institute within the framework of the project "Methodology for Automatic Identification and Classification of Mammals in Trap Camera Images" funded by COLCIENCIAS.

Built With

• matlab